DEFENCE+ industry
Impact of exponential growth
laws in military systems
Dr Carlo Kopp
‘
An arguably bizarre situation
has arisen through an irrational
belief in the management
circles of many organisations
that exponentially growing
IT/ICT obviates the need
for smart, well educated
and trained personnel.
’
Intel Core i7 Nehalem quad core processor chip die.
A DEFINING feature of the
contemporary military computing
environment is the direct and indirect
impact of exponential growth laws,
the best known of which is Moore’s
Law. Defined by Intel co-founder
Gordon Moore in a 1965 paper,
this law predicts that the density of
computer chips will double every
18 to 24 months. Despite repeated
predictions that Moore’s Law would fail
sooner or later, nearly a half century
later it continues to hold. Whenever
a technological obstacle appeared,
engineers devised a way around it.
36
Positive Impacts of exponential
growth laws
There’s no doubt exponential growth laws have
produced valuable gains in military information
processing and collection applications, also in
military networking and sensor technologies. A
contemporary radar system, electro-optical system
or networking terminal will be more reliable, more
sophisticated functionally, and often much more
capable than its predecessors of one, two or more
decades ago.
Abundant computing power has been especially
valuable in military applications such as sensors,
networking, navigation, embedded system
management processing, weapon guidance,
mission planning systems, simulators and
databases. The declining cost of computing power
has been a major enabler in many key areas,
putting computationally demanding problems
within easy reach.
Two good examples are Space Time Adaptive
Processing (STAP) in radar, used to reject ground
clutter, and Orthogonal Frequency Division
Multiplexing (OFDM) used in commercial and
military digital communications. STAP is now being
deployed in a number of radar designs, Western
and Russian, to improve the capability to detect
small low flying targets, yet the technique was
understood but out of reach for well over a decade,
as computers of the time were not powerful
enough to perform the computations in real time.
OFDM is now the mainstay of broadband and digital
television modulation, yet when it was devised four
decades ago, only the fastest supercomputers of
the day could perform the necessary calculations
in real time.
Exponential growth in the performance of optical
fibre communications has produced important
gains in fixed military infrastructure bandwidth,
but also internal cabling bandwidth in platforms
such as aircraft or warships, where internal copper
cabling historically often set hard performance
limits on bandwidth between avionic or combat
system components.
Vastly improved density and reliability in hard
disks is seeing increasingly their use in embedded
applications, but has also been of pivotal importance
as an enabling technology in areas such as GIS,
simulations and data mining. Petabyte databases
for storing digital imagery, 3D GIS mapping data,
publications and other high value data are now
sufficiently affordable to permit much wider use of
these valuable technologies.
Improvements in digital imaging chips have enabled
important growth in imaging sensors where these
operate in the visible colour bands, and can thus
exploit exponential growth in commercial imaging
products. Recent technology demonstrations such
as the DARPA 1.8 Gigapixel ARGUS-IS imaging
sensor would be simply infeasible without current
commodity commercial digital imaging chips.
Commercial high definition CCD or CMOS imaging
chips have been used in reconnaissance and
DEFENCETODAY
DT_JUN2011.indd 36
31/05/11 2:11 PM
surveillance imaging products for almost a decade.
Where an exponential growth law applies to a basic
technology used or usable in a military system, as
long the growth law holds, there will be a sustained
period of improvement in the resulting capability.
How useful that improvement is depends critically
upon the capability in question, and what other
factors influence or limit capability growth.
Negative impacts of exponential
growth laws
While exponential growth laws have produced
enormously valuable gains in many military
applications, they have also produced numerous
undesirable negative impacts, often producing
disproportionate damage. A philosophical
observation is that the technology itself knows nor
cares about how it is used or abused. A related
observation is that this reality has become the
basis upon which anti-technology Luddites most
typically argue against new technology.
The most widely accepted negative impact of
exponential growth laws is early obsolescence
of those technologies growing exponentially, as
production life becomes ever shorter. This results
in many military systems being in a permanent
state of being upgraded, as components become
unsupportable due to obsolescence of parts used
to build them. While this can be planned for, the
human dimension of the problem frequently results
in a failure to do so, with resulting life cycle cost
blowouts.
More importantly, the single biggest negative impact
produced as a byproduct of exponential growths
laws is a completely irrational and widely held
belief amongst technologically illiterate managers,
bureaucrats and planners that exponential growth
laws apply universally to all technologies, and
especially all military technologies. As a result,
there has been a widespread abandonment of
traditional physics and operational analysis based
military and strategic planning in Western nations,
as it is somehow believed that exponential growth
laws will overcome the laws of physics and basic
mathematics in all other areas.
Yet the material reality is that exponential growth
laws cannot apply to any technologies in the
‘kinetic domain’ involving mechanical forces and
physical mass in objects, and in the ‘information
domain’ exponential growth laws simply do not
apply to radio-frequency or optical apertures, as
they do not apply to software algorithms.
The irrational belief in the ‘universality’ of
exponential growth laws is paralleled by a related
and no less irrational belief that the ever declining
cost of exponentially growing basic technologies
must be paralleled by an ever declining cost in all
other technologies. Yet as empirically observable
reality shows, the declining cost rule does not
and indeed cannot apply either to ‘kinetic domain’
technologies, or non-exponential ‘information
domain’ technologies such as radio frequency or
optical apertures, or software algorithms.
One might argue that this negative impact is
indeed not the fault of the technology but of the
organisations that choose to employ technologically
illiterate personnel in jobs demanding high
levels of technological literacy. This argument
inevitably leads to a chicken versus egg polemic
over advanced IT/ICT equipment enabling the
replacement of technologically literate personnel
with illiterate personnel.
Technologically highly literate personnel are today
much less employable than a decade ago, yet
are far more needed now than ever before. An
arguably bizarre situation has arisen through an
irrational belief in the management circles of many
organisations that exponentially growing IT/ICT
obviates the need for smart, well educated and
trained personnel.
A closely related problem is ‘information overload’
or ‘Information Fatigue Syndrome’, a now pandemic
problem in developed nations characterised by
management personnel at all levels being unable
to make rational decisions as they try to cope with
incessant bombardment with erroneous, deceptive
or simply irrelevant data.
The latter itself is another most interesting and
both unexpected/unintended consequence
of an exponentially growing ability to digitally
collect, store, process and rapidly disseminate
vast volumes of data across military and civilian
organisations and communities, nations and indeed
the international community. The technology is
transparent to the data being distributed, so fact
or fiction, the latter intended or unintended, can
be propagated at unprecedented speeds. The
sad reality is that far more fiction is propagated
than fact across the global IT/ICT infrastructure,
evidenced by the less than palatable reality that
often more than 90 per cent of electronic mail
traffic entering mail server computers is unwanted
spam.
The increasing dependency of developed nations
upon pervasive digital technology creates in itself
a single point of failure vulnerability, as traditional
human processing and handling of information and
data is abandoned, and skills to do so lost and
no longer taught. Whether damage is inflicted by
human driven effects such as cyber warfare or
electro-magnetic weapons or by ‘Mother Nature’
solar flares, the material reality is that exponential
growth is creating in turn a growing vulnerability to
the loss of the IT/ICT infrastructure, regardless of
the cause of the loss.
Exponential Growth Laws
and their Limitations
In assessing the impact of exponential laws on
military and commercially developed systems an
understanding of the relative exponential laws and
their historical impact on technological capability
and performance is essential.
Moore’s Law is one of a family of growth laws
identified in recent decades, which define
performance, capability in a range of different
technologies.
Exponential growth laws are now producing
pronounced effects, especially in any technologies
where a large consumer market can support
the research and development costs. Computer
hardware of all persuasions, digital networking
running over optical fibres and digital camera
chips, have advanced dramatically in performance
This Curtiss-Wright Controls SVME/DMV-1905
6U VME board with an Intel Core i7 processor is
typical of state of the art embedded hardware,
which permits the insertion of unprecedented
computing power into military systems.
This state of the art Seagate Momentus XT
hybrid hard disk drive is an interesting example
of Kryder’s exponential growth law in its 500
Gigabyte magnetic disk technology, Moore’s
exponential growth law in its 4 Gigabyte DRAM
cache, and mechanically limited head technology
with a 4 millisecond access time, not significantly
better than a 3.5 inch drive of a decade ago.
DEFENCETODAY
DT_JUN2011.indd 37
37
31/05/11 2:11 PM
DEFENCE+ industry
and capability over the past decade, a trend set to
continue.
But many other technologies are not advancing
exponentially even if there is a widely held
expectation outside the science and engineering
communities that these somehow should do so.
Nature being what it is, some things can happen
and others cannot.
Computer software is an area where the opposite
trend has emerged, where ever increasing
complexity of software systems often exceeds the
gains of Moore’s Law, resulting in a nett decrease
in the performance of an application running on
a piece of computing hardware. This observation
was first attributed to famous computer scientist
Niklaus Wirth, which essentially says that “software
is getting slower more rapidly than hardware
becomes faster.”
A major limitation arising in computer systems is
Amdahl’s Law of parallel processing. Increasingly,
we see a drift from computer chips with single
processors (CPU) to multiple core designs, with
two, four, six or more ‘cores’ on a single chip.
Recent prototype chips have run to dozens or more
cores. The idea is that many computing problems
can be partitioned into multiple parallel problems,
which can be then computed independently on
multiple cores. Double the number of cores, double
the speed, quadruple the number of cores, and
quadruple the speed. While multiplying the number
of cores can indeed produce a proportionate
increase in computational speed for many problems,
this is not generally true. Amdahl’s Law states that
speed-up depends more critically on what parts
of the software have mutual dependencies that
preclude concurrent computation than it does on
the number of processing cores. For many basic
computing problems, having a hundred processing
cores may yield little speed advantage over the use
of a single core.
The growth law for hard disks expresses limitations
in exponential performance improvement. While
the density of information storage on hard disks
has increased markedly over the past decade
the access time for data on hard disks has only
improved two to three-fold in 20 years. Access
time is determined by the mechanical rotational
velocity of the disk and the time to mechanically
move the magnetic transducer head over the
rotating disk platter. So the time it takes to find a
block of data on a current Terabyte disk drive may
Optical imaging chips may well be growing in resolution
exponentially, but the optical apertures used, such as
lenses, do not. While the professional lens on this digital
camera outperforms the 12.3 Megapixel camera body,
it would be inadequate in sharpness for a future 120
Megapixel camera body.
38
be 50 per cent shorter than what it took fifteen
years ago to locate a block on a 300 Megabyte
period disk.
Wireless radio frequency link performance is
driven by the Friis Inverse Square Law, and by
antenna performance by the ratio of antenna size
to operating wavelength, both as a result producing
little if any evolution in many respects since radio
technology was introduced a century ago. A 10
GigaHertz band antenna dish with a gain of 20
deciBels will be no different in size today than
it was 50 years ago. This hard physics limit in
antenna technology is thus a hard constraint not
only on radio networks but also on radars, radiofrequency and laser beam weapons, and any other
technology which relies upon the transmission
of a signal or power over a free space or an
atmospheric path.
Another hard limit on achievable bandwidth in
wireless radiofrequency channels, whether
datalinks or radar, is spectral congestion, arising
from ever increasing demands for spectrum
used in mostly consumer applications. This has
reached the point of forcing early obsolescence
of military equipment such as radar and satellite
links, where new equipment operating in different
bands must be deployed. A more insidious problem
is interference from adjacent bands, where cheap
commercial equipment leaks transmissions into
frequencies in which they do not belong. The latter
is beginning to impact satellite navigation systems,
and is a well established problem in shared
consumer radio frequency bands.
Exponential density increases in monolithic
microwave chips have not resulted in exponential
growth in the performance or other capabilities
of microwave electronics, although reliability
has improved vastly, and cost has declined
dramatically. This is because such chips handle
microwave signals, and propagation physics in
internal and external interconnections become a
major limitation to performance growth. Power
handling capability, i.e. on-chip cooling, has limited
performance growth in computer chips, and also
often sets hard limits on performance in microwave
chips.
Limitations to exponential growth also apply to the
performance of digital imaging systems, using an
optical lens or mirror to focus an image on to a CCD
or CMOS technology imaging chip, as observed
in digital cameras, webcams, camcorders and
cellular phones. We are now at the point where
the spatial bandwidth, a measure of how much
information such a system can capture in an
image, is mostly limited by the sharpness of
the lens or mirror optics, not the number of
Megapixels in the imaging chip. While professional
camera lenses are still ahead of imaging chips
in bandwidth performance, in consumer devices
the imaging chip bandwidth outstripped lens
performance some years ago, with consumers
paying for Megapixels that are effectively unused.
A 10 Megapixel imaging chip with a suitable
digital processor can today outperform all but the
largest format wet film imaging technologies, in
dynamic range, colour accuracy, and noise (grain)
impairments.
The vast improvements seen in visible band
imaging chips have not been observed in military
thermal imaging chips, in part due to the difficulties
in manufacturing with exotic materials, and in part
due to the high development costs during a period
of collapsing military research and development
budgets.
Conclusion
The pervasive nature of modern digital technology
has produced some genuine and deep problems
at the human end of the equation, mostly as
byproducts of a prior failure to properly educate,
train and staff organisations, itself a result of a
failure to understand or anticipate the impact of
exponential growth laws. Clearly most of these
problems can be corrected by properly staffing
organisations with personnel who are trained to
think critically and understand the limitations of
exponentially growing technologies. The biggest
challenge is not however in the implementation of
corrective action, but in gaining acceptance that
action is even required!
Antennas, whether used for radar or communications, neither shrink nor improve in performance exponentially, as they
are limited in performance growth by radio physics and the cooling capability of the carrying platform.
DEFENCETODAY
DT_JUN2011.indd 38
31/05/11 2:11 PM